Purpose: Radiation doses to cardiac substructures from radiotherapy (RT) of non-small-cell-lung-cancers (NSCLC) have been linked to post-RT cardiac toxicities. Given a large number of cardiac substructures and surrounding organs-at-risk (OARs), it can be challenging to optimize a desirable treatment plan due to tradeoffs between OARs sparing and target coverage. We built a cardiac-substructures-based KBP (CS-KBP) model and tested its performance against cardiac-based KBP (C-KBP) model and manually optimized treatment plans.
Methods: CS-KBP/C-KBP models were built with 28 previously-treated NSCLC patients. Both models were trained with plans that preferentially spare heart. While the C-KBP training plans were created with whole heart structures, the CS-KBP model training plans each has fifteen cardiac substructures (coronary arteries, valves, great vessels, and chambers of the heart). CS-KBP training plans reflect cardiac-substructure sparing preferences. We evaluated our models on 29 additional patients. Three sets of treatment plans were generated: (1) manually-optimized, (2) C-KBP model-generated, and (3) CS-KBP model-generated. Plans were normalized to receive the prescribed dose to at-least 95% of the target. Two-tailed paired-sample t-test was performed for clinically relevant dose-volume-metrics to evaluate the performance of CS-KBP model against C-KBP model and clinically plans, respectively.
Results: Overall results show significantly improved (P <0.01) cardiac substructures sparing by CS-KBP in comparison to C-KBP and clinical plans. Average left anterior descending artery volume receiving 15Gy (V15Gy) was 1.23±1.76cc, 1.05±1.68cc, and 0.69±1.57cc by clinical, C-KBP, and CS-KBP plans, respectively. Similar trend was observed for other clinically relevant dose-volume metrics. CS-KBP, however, increased doses to lungs (D35%=13.65±5.84Gy vs 12.76±5.38Gy and 12.47±4.46Gy for C-KBP and clinical plans, respectively) without exceeding its standard tolerance.
Conclusion: Our CS-KBP model significantly improved cardiac-substructure sparing without exceeding tolerances of other OARs and compromising target coverage. This model may offer reduced planning time, improve plan quality, and can be used to drive patient outcomes.
Funding Support, Disclosures, and Conflict of Interest: Supported in part by Winship Cancer Institute # IRG-21-137-07 -IRG from the American Cancer Society
TH- External Beam- Photons: Treatment planning using machine learning/Knowledge Based Planning/automation